1. Field of the Invention
This invention relates to the field of intelligent networks that include connection to the physical world. In particular, the invention relates to providing distributed network and Internet access to sensors, controls, and processors that are embedded in equipment, facilities, and the environment.
2. Description of the Related Art
Sensor networks are a means of gathering information about the physical world and then, after computations based upon these measurements, potentially influencing the physical world. An example includes sensors embedded in a control system for providing information to a processor. The Wireless Integrated Network Sensor (WINS) development was initiated in 1993 under Defense Advanced Research Projects Agency (DARPA) program support. The Low-power Wireless Integrated Microsensors (LWIM) program pioneered the development of WINS and provided support for the development of fundamental low power microelectro-mechanical systems (MEMS) and low power electronics technology. The LWIM program supported the demonstration of the feasibility and applicability of WINS technology in defense systems. See: K. Bult, A. Burstein, D. Chang, M. Dong, M. Fielding, E. Kruglick, J. Ho, F. Lin, T.-H. Lin, W. J. Kaiser, H. Marcy, R. Mukai, P. Nelson, F. Newberg, K. S. J. Pister, G. Pottie, H. Sanchez, O. M. Stafsudd, K. B. Tan, C. M. Ward, S. Xue, J. Yao, “Low Power Systems for Wireless Microsensors”, Proceedings of International Symposium on Low Power Electronics and Design, pp. 17–21, 1996; J. G. Ho, P. R. Nelson, F. H. Lin, D. T. Chang, W. J. Kaiser, and O. M. Stafsudd, “Sol-gel derived lead and calcium lead titanate pyroelectric detectors on silicon MEMS structures”, Proceedings of the SPIE, vol. 2685, pp. 91–100, 1996; D. T. Chang, D. M. Chen, F. H. Lin, W. J. Kaiser, and O. M. Stafsudd “CMOS integrated infrared sensor”, Proceedings of International Solid State Sensors and Actuators Conference (Transducers '97), vol. 2, pp. 1259–62, 1997; M. J. Dong, G. Yung, and W. J. Kaiser, “Low Power Signal Processing Architectures for Network Microsensors”, Proceedings of 1997 International Symposium on Low Power Electronics and Design, pp. 173–177, 1997; T.-H. Lin, H. Sanchez, R. Rofougaran, and W. J. Kaiser, “CMOS Front End Components for Micropower RF Wireless Systems”, Proceedings of the 1998 International Symposium on Low Power Electronics and Design, pp. 11–15, 1998; T.-H. Lin, H. Sanchez, R. Rofougaran, W. J. Kaiser, “Micropower CMOS RF components for distributed wireless sensors”, 1998 IEEE Radio Frequency Integrated Circuits (RFIC) Symposium, Digest of Papers, pp. 157–60, 1998; (Invited) G. Asada, M. Dong, T. S. Lin, F. Newberg, G. Pottie, H. O. Marcy, and W. J. Kaiser, “Wireless Integrated Network Sensors: Low Power Systems on a Chip”, Proceedings of the 24th IEEE European Solid-State Circuits Conference, 1998.
The first generation of field-ready WINS devices and software were fielded in 1996 and later in a series of live-fire exercises. The LWIM-II demonstrated the feasibility of multihop, self-assembled, wireless network nodes. This first network also demonstrated the feasibility of algorithms for operation of wireless sensor nodes and networks at micropower level. The original WINS architecture has been demonstrated in five live fire exercises with the US Marine Corps as a battlefield surveillance sensor system. In addition, this first generation architecture has been demonstrated as a condition based maintenance (CBM) sensor on board a Navy ship, the USS Rushmore.
Prior military sensor systems typically included sensors with manual controls on sensitivity and radio channel selection, and one-way communication of raw data to a network master. This is wasteful of energy resources and inflexible. In the LWIM network by contrast, two-way communication exists between the sensor nodes and the master, the nodes contain signal processing means to analyze the data and make decisions on what is to be communicated, and both the communications and signal processing parameters can be negotiated between the master and the sensor nodes. Further, two-way communications enables consideration of more energy-efficient network topologies such as multi-hopping. The architecture is envisioned so that fusion of data across multiple types of sensors is possible in one node, and further, so that the signal processing can be layered between special purpose devices and the general-purpose processor to conserve power. The LWIM approach to WINS represented a radical departure from past industrial and military sensor network practice. By exploiting signal processing capability at the location of the sensor, communications energy and bandwidth costs are greatly reduced, allowing the possibility of scalably large networks.
The DARPA sponsored a second program involving both UCLA and the Rockwell Science Center called Adaptive Wireless Arrays for Interactive Reconnaissance, surveillance and target acquisition in Small unit operations (AWAIRS), whose genesis was in 1995. Its focus has been upon the development of algorithms for self-assembly of the network and energy efficient routing without the need for masters, cooperative signal processing including beamforming and data fusion across nodes, distributed self-location of nodes, and development of supporting hardware. A self-assembling network has been demonstrated. Moreover, the AWAIRS program includes notions such as layered signal processing of signals (including use of multiple processors within nodes, as in LWIM), and data aggregation to allow scaling of the network. A symposium was held in 1998 to discuss the implications of such sensor networks for a wide variety of applications, including military, health care, scientific exploration, and consumer applications. The AWAIRS nodes have also been used in condition based maintenance applications, and have a modular design for enabling various sensor, processing, and radio boards to be swapped in and out. There is now a confirmed set of WINS applications within the Department of Defense for battlefield surveillance and condition based maintenance on land, sea and air vehicles, and WINS technology is being considered as a primary land mine replacement technology. See: J. R. Agre, L. P. Clare, G. J. Pottie, N. P. Romanov, “Development Platform for Self-Organizing Wireless Sensor Networks,” Aerosense '99, Orlando, Fla., 1999; K. Sohrabi, J. Gao, V. Ailawadhi, G. Pottie, “A Self-Organizing Sensor Network,” Proc. 37th Allerton Conf. on Comm., Control, and Computing, Monticello, Ill., September 1999; University of California Los Angeles Electrical Engineering Department Annual Research Symposium, 1998; K. Yao, R. E. Hudson, C. W. Reed, D. Chen, F. Lorenzelli, “Blind Beamforming on a Randomly Distributed Sensor Array System,” IEEE J. Select. Areas in Comm., vol. 16, no. 8, October 1998, pp. 1555–1567.
There are also a number of commercial sensor technologies that are related to WINS, in that they include some combination of sensing, remote signal processing, and communications. Some of these technologies are described herein, along with some expansion upon the specific features of LWIM and AWAIRS.
FIG. 1 is a prior art control network 100. The network 100 typically includes sensors 102, a master 104, and possibly a plurality of actuators 106 that are tightly coupled, a configuration that results in a low delay in the feedback loop. Typically, the sensors 102 have parameters that are controlled by the master 104. The network may include a number of controllers and actuators. Results of actuation are detected by the sensors 102, which, together with the logic in the master 104, results in a control loop. Typically, raw measurements are forwarded to the master 104 with little or no processing (e.g., low pass or passband filtering). The master 104 reports the results to a computer network 108. Furthermore, the master 104 accepts new programming from that network 108.
FIG. 2 is a prior art sensor network 200. The typical network includes a number of sensor nodes 202, a master 204, and a user interface 206. The master 204 is often just another sensor node, or may be a more sophisticated device. The elements of the network 200 are hand registered, and there is limited self-assembly and reconfiguration capability residing in the network 200 (e.g., updating of addresses as new nodes are registered). Typically, the parameters of the sensors 202 are controlled by the master 204, and raw measurements are forwarded to the master 204 with little or no processing. For example, in remote meter reading applications the meter value at some particular time is sent. However, in LWIM networks extensive processing is performed to make decisions, and thus reduce the communications traffic and relieve the burdens of the master. The master 204 reports the results to the user interface 206, following some computation, using a long range communication link 208. The limitation that inheres is that the interface 206 allows for downloading of new programming (for example, on a laptop computer) via the master 204. In a typical military or meter-reading system however, there is only one-way communication upwards from the sensor 202 to the master 204, and thus no tuning of node parameters is possible.
FIG. 3 is a prior art AWAIRS sensor network 300. The sensor nodes 302 of the AWAIRS network 300 include extensive signal processing in order to reduce communications. The sensor nodes 302 can include multiple processors of differing types, and can progress through several levels of signal processing in performing target detection and identification. The sensor nodes 302 can also include ranging devices for position location. Moreover, the sensor nodes 302 enable cooperative behaviors such as data fusion, beamforming, and cooperative communications. The network 300 is self-organizing, and will establish routing to minimize energy consumption. Multihop routing is supported. The network 300 does not require long-range links, but can include them, and may directly connect to a computer and user interface 306. Moreover, the sensor nodes 302 may interact with a number of user interfaces 306. Data aggregation may be included in a path from the remote sensors to an end destination.
FIG. 4 is an example of a prior art sensor network 400 using distributed signal processing. Source 1 emits a signal that is detected by sensors 1, 2, and 3. Sensor node 1 can become designated as a fusion center to which some combination of data and decisions are provided from sensor nodes 2 and 3. Sensor node 1 then relays the decision towards the end user using a specific protocol. Source 2 emits a signal that is detected by sensor node 4. Sensor node 4 performs all processing and relays the resulting decision towards the end user.
Sensor node 6 receives the signals emitted by both sensors 1 and 4. Sensor node 6 may pass both decisions or perform some further processing, such as production of a summary activity report, before passing information towards the end user. The end user may request further information from any of the sensor nodes involved in processing data to produce a decision.
FIG. 5 is an example scenario for self-organization in a prior-art sensor network such as AWAIRS. In the limit of short hops the transceiver power consumption for reception is nearly equal to that of transmission. This implies that the protocol should be designed so that radios are off as much of the time as possible, that is, the Media Access Controller (MAC) should include some variant of Time-Division Multiple Access (TDMA). This requires that the radios periodically exchange short messages to maintain local synchronism. It is not necessary for all nodes to have the same global clock, but the local variations from link to link should be small to minimize the guard times between slots, and enable cooperative signal processing functions such as fusion and beamforming. The messages can combine health-keeping information, maintenance of synchronization, and reservation requests for bandwidth for longer packets. The abundant bandwidth that results from the spatial reuse of frequencies and local processing ensures that relatively few conflicts will result in these requests, and so simple mechanisms can be used.
To build this TDMA schedule, the self-organization protocol combines synchronism and channel assignment functions. It supports node-to-node attachment, node-to-network attachment, and network-to-network attachment. The distributed protocol assigns progressively less of the TDMA frame to invitations and listening as the network becomes more connected. The result is contention-free channel assignments for the sensor nodes in a flat (peer-to-peer) network, where the channels consist of some combination of time and frequency assignments. Invitation slots are allocated even when the network is mature to allow for reconfiguration.
Upon construction of the set of links, the routing is then built. If the nodes are powered by batteries, the network will have a life-cycle which begins in a boot-up, proceeds through a phase of maximum functionality, decline, and finally failure. Every bit that is exchanged hastens the end of the network. Particular nodes may be more heavily stressed by traffic than others (e.g., those in the vicinity of a gateway or other long-range link). Thus, routing protocols must to some extent be energy-aware, to sustain useful operation as long as possible. The minimum energy path is not necessarily the most desirable. Rather routes are ordinarily chosen to extend operation, although high priority messages may be routed for low latency, even if this exhausts precious network resources. The predictability of flow to and from a relatively small number of gateways enables infrequent construction of sets of paths to these data points, minimizing overhead.
FIG. 6 is an example scenario of self-location in a prior art sensor network. In this scenario, sensor nodes 2, 5, 8, and 9 contain an absolute position reference mechanism, for example Global Position System (GPS) or hand registration of position. Furthermore, all sensor nodes include transducers and receivers for radio frequency (RF) or acoustic ranging. As such, the network elements are homogeneous except possibly for nodes 2, 5, 8, and 9, as these nodes provide position and timing reference. Sensor node algorithms estimate ranges to neighboring sensor nodes using a time difference of arrival (TDOA) scheme. The results are used to set up either linear or non-linear systems of equations using either distributed or centralized algorithms. For example, if all nodes can hear the four references, standard GPS algorithms can be used independently by each node. If nodes can only hear near neighbors, iterative procedures may be employed. Using this system, a position determination is made when a sensor node hears at least four neighboring sensor nodes. While four nodes are required for an absolute position determination in a three-dimensional system, results are better when more than four nodes are detected. Also, only a small percentage of the sensor nodes of a network are required to make an absolute position determination.
FIG. 7 is an example of sensor/internet connections in a prior art sensor network 700. The sensor nodes 702 may be cameras, interfaced to a computer by means of an electronic card. The interface card 704 allows for control by a computer 706 of a limited number of parameters. The network interface 708 includes, for example, a modem card in a computer, telephone line access, or access to an Internet Service Provider (ISP). The images processed by the host computer 706 can be viewed remotely by users with similar Internet access, for example when the images are placed on a publicly available World Wide Web (web) site. The images placed on a web site may be downloaded and modified using remote computers 714 and interfaces 712 with web site access. While this network makes use of standard software, it requires an expensive interface between the computer 706 and each sensor node 702. Furthermore, manual configuration of the connection and the software is typically required to attach each sensor node 702 to the network 710.
As another example, in a prior art system designed for airport security, a seismic sensor and energy detector circuit is used to trigger a digital camera under the control of a computer. The image and seismic record are conveyed by wireless means to another computer, and from there posted to a web site. The trigger level can be controlled remotely via the web site. However, only one remote unit is supported, with no networking of multiple sensors, and with the requirement of a costly interface platform, or computer, at both ends.
While these examples indicate some aspects of wireless sensor network technology, many desirable features are absent. Each of these systems either lacks ease of use, ability to use standard development tools to extend them, and/or ability to operate in variable or hostile environments. For example, the wireless communications technique may be vulnerable to jamming or interference, or the platform may consume too much energy for long-term remote operation, or it may lack simple connectivity to the Internet or support for database services, or only support a limited number of sensing modes.
Wireless network technology has progressed so that the WINS platform or set of platforms is required to support standard operating systems and development environments, and be capable of being easily integrated into larger networks. Only in this fashion can the physical world be seamlessly connected to the many resources available through the Internet and other networks. In particular, the WINS platforms are required to provide a familiar and convenient research and development environment. The cumbersome embedded systems of past implementations are not appropriate for this next generation of progress. The custom operation systems developed for past generations of low power sensor nodes have an inconvenient development environment and are not supported by the familiar, high productivity, powerful, development tools needed by the research and development community. Furthermore, conventional approaches would yield a system where a platform operating with a conventional embedded operating system would require excessive operating power. This prevents developers from facing and solving the challenges of low power system design. The development of these essential capabilities requires a fundamentally different WINS node and network architecture.